TasteGraph Sentinel — Autonomous Qloo Cultural Intelligence Agent

Elevator Pitch

An autonomous AI agent grounded in Qloo's 250M+ cultural entity graph that maps cross-domain taste affinities across Cinema, Music, Gastronomy, Podcasts, Literature, and Fashion, eliminating cultural hallucination in AI agents.

Inspiration

When you ask standard LLMs for lifestyle, travel, or media recommendations based on your tastes, you inevitably receive generic, cliché answers. An LLM prompted with "What music and dining matches someone who loves Blade Runner 2049?" will hallucinate arbitrary popular spots or generic synthwave tracks. LLMs simply lack empirical taste correlation.

Qloo solves this with over 250 million real-world cultural entities and billions of taste affinities. We wanted to build an autonomous agent that directly taps into this cultural nervous system, grounding AI reasoning in authentic human taste.

What It Does

TasteGraph Sentinel is an autonomous cultural reasoning agent and interactive Web Studio:

1. Multi-Domain Taste Projection: Takes any cultural seed (a film like Interstellar, a musician like Vangelis, or an aesthetic like Mid-Century European) and autonomously expands it across Cinema, Music, Dining, Podcasts, Literature, and Fashion.
2. Cultural Coherence Scoring: Computes a mathematical resonance metric across disparate domains, ensuring that high-conviction affinities remain harmonically aligned.
3. Curated Experiential Itineraries: Translates multi-domain affinities into an actionable day plan (Morning Routine, Afternoon Flow, Evening Sensory, Nightcap Reflection).
4. Resilient Dual-Mode Execution: Integrates directly with Qloo REST API v2 while providing curated offline fallback knowledge for zero-latency CLI and offline agent autonomy.

How We Built It

• Core Agent Architecture: Built in Python with modular layers (QlooClient, TasteVectorEngine, and CulturalAgent).
• Taste Affinity Scoring: Implemented mathematical vector scoring and multi-domain projection algorithms in affinity.py.
• Automated Verification: Engineered a complete pytest suite (9/9 passed in 0.32s) verifying live API calls, offline fallbacks, and markdown report generation.
• Interactive Web Studio: Developed a lightweight, zero-dependency Web Studio in docs/index.html featuring interactive archetype resolvers, domain filter tabs, and one-click JSON/Markdown dossier export.
• Continuous Deployment: Hosted the interactive demo live on GitHub Pages.

Challenges We Overcame

Mapping cross-domain entities without falling into circular or overfit clusters required designing the Cultural Coherence Metric, which rewards diversity across distinct artistic mediums while penalizing out-of-character outliers. Additionally, ensuring the agent functions both online (with live Qloo API keys) and offline (for zero-latency local agentic pipelines) required building an archetype-anchored fallback architecture.

Accomplishments We're Proud Of

• 100% automated test coverage across all client, scoring, and agent orchestrator workflows in 0.32s.
• Clean, responsive Web Studio deployed on GitHub Pages with zero external frontend framework bloat.
• Grounded intelligence that immediately provides tangible, delightful cultural recommendations that feel deeply human and curated.

What We Learned

Empirical cultural data transforms AI from a generic conversationalist into a truly sophisticated personal curator. Connecting LLM reasoning to Qloo's Taste Graph proves that the future of agentic AI lies in grounding models with real human taste networks.

Built With

• Python 3.10+
• Qloo REST API v2
• Pytest
• GitHub Pages
• Vanilla HTML5 / Modern CSS

Try It Out

• GitHub Repository: https://github.com/bhomvratrai-glitch/qloo-cultural-agent
• Live Interactive Demo: https://bhomvratrai-glitch.github.io/qloo-cultural-agent/

Built With

• agentic-ai
• cultural-intelligence
• github
• pytest
• python
• qloo-api